Content-based Copy Detection Task

This year, we participated in TRECVID 2010 content-based copy detection task. In this notebook paper we will describe our work in details. Different from last year when we just used SURF (speeded up robust feature) as visual feature, this year we employed a combination of four different features for our rough detection process: global SURF, center SURR, global color correlogram and center correlogram. What highlights our work is that , to achieve high accuracy detection ,we proposed a method using Non-orthogonal Binary Subspace in our accurate detection process. Finally we submitted 4 different runs in CBCD task, their description are as follows: 1.NJU.m.nofa.norank1:pre-process+feature extraction+search+post process 1 2.NJU.m.balanced.rank1:pre-process+feature extraction+search+post process 2 3.NJU.m.nofa.comp2:pre-process+feature extraction+search+post process 3 4.NJU.m.balanced.comp2:pre-process+feature extraction+search+post process 3 The only difference among these runs is that when producing the final result, we adpoted different post process methods.